Some portions of the current documentation page appear to be outdated or unclear. Here I propose changes to address issues I have found.

• Standardise documentation to use only Data Owner instead of both Data Provider and Data Owner. For example, this section uses Data Owner in the Figure, despite the section being named Data Providers.

• In the Setup section for the local server, instruct users to use the virtual environment created in the Installation section, if they made one.

• Add (conditional?) logout steps to the setup_benchmark_tutorial.sh, setup_data_tutorial.sh, setup_model_tutorial.sh and simulate_data_association_approval.sh scripts so that their login steps work properly. Currently, the logins fail if logged into a different user, which may happen for people doing multiple tutorials in a row.

• Fix the final rm -rf command in the Clean Up sections. The ~/.medperf/localhost_8000 directory mentioned does not exist. Rather, there are multiple ~/.medperf/<some_dir>/localhost_8000 directories to be removed instead. There is also a ~/.medperf/demo directory that I believe is also related to tutorials and should also be removed.

• In Section 2 of the Benchmark Committee section, explicitly mention that the workspace is the medperf_tutorial directory to avoid confusion, similarly to what is done in Section 4

• In Section 5 of the Benchmark Committee, the packaging script fails for the Data Preparation MLCube because the mlcube.yaml file does not have the report_file fields. It appears to be outdated.

  • If this error is ignored by the user, they can follow along the rest of the tutorial. The provided hosted MLCube (on DockerHub) is built correctly and compatible with the provided mlcube.yaml file. Therefore, the submit step as presented in the current documentation works properly.

  • However, if the local mlcube.yaml file is updated to include the required report_file fields, the packaging script will work but submission will fail, as the image on DockerHub does not support a report_file input. So it would need to be updated too.

  • Another option: make the report_file field optional in the Schema.